An empirical comparison of the performance of alternative option pricing models

This paper presents a comparison of alternative option pricing models based either on jump-diffusion nor stochastic volatility data generating processes. We assume either a smooth volatility function of some previously defined explanatory variables or a model in which discrete-based observations can be employed to estimate both path-dependence volatility and the negative correlation between volatility and underlying returns. Moreover, we also allow for liquidity frictions to recognize that underlying markets may not be fully integrated. The simplest models tend to present a superior out-of sample performance and a better hedging ability, although the model with liquidity costs seems to display better in-sample behavior. However, none of the models seems to be able to capture the rapidly changing distribution of the underlying index return or the net buying pressure characterizing option markets.